OpenAI limits GPT-5.6 rollout after U.S. government security request
OpenAI restricted GPT-5.6 access to 20 partners, with broader availability expected by July 2026, as U.S. security-driven limits on frontier AI models fuel debate over competitiveness, cybersecurity and Chinese open-weight rivals.
OpenAI said it would limit the rollout of its GPT-5.6 models following a U.S. government request, with reports saying access was initially restricted to 20 partners ahead of a broader public launch expected by July 2026. The move, alongside U.S. actions affecting Anthropic models, has intensified debate over whether security reviews and export-related restrictions on leading domestic AI systems could slow commercialization, alter cybersecurity planning and give Chinese competitors using cheaper open-weight models more room to catch up.
- open-weight models: AI models released with downloadable parameters that users can run on their own infrastructure
- export controls: Government restrictions on the sale or transfer of sensitive technologies across borders
- restricted rollout: A limited release of a product to a small group of partners before wider availability